Principal Investigator: Paul Gader
Start Date: January 15, 2019
End Date: August 15, 2020
The goal of this project is to use Machine Learning to characterize the response of underwater fauna to disturbances using multi-channel underwater acoustic and video data. Acoustic processing will focus on using a sequence consisting of an anomaly detector followed by a classifier. A class of algorithms called Dirichlet Process Mixture Models will be investigated for anomaly detection. Hidden Markov Models will be investigated for use as classifiers. Video processing will consist fish tracking followed by extracting features from the fish tracks and the images of the fish. The project will also include participation in data collections and on-site algorithm demonstrations.